Machine-learning-based beam steering in a hybrid plasmonic nano-antenna array

Abstract

In this paper, different beam-steering techniques are introduced using an 8 × 8 hybrid plasmonic nano-antenna array operating at 1550 nm. Two conventional techniques of the switched-beam and phased array are considered for implementing beam steering. For a switched-beam antenna, the beam steering is achieved by switching between the feed antenna elements, whereas the feed antenna array is attached to the back surface of a fishnet achromatic-metalens to collimate the rays. In the phased array antenna, the beam is steered by estimating the appropriate feeding phases of the 64 elements using a deep neural network (DNN) either with or without a lens and comparing the results with those obtained using a traditional method. Finally, a hybrid technique based on activating only a subset of antenna elements in the existence of the lens is proposed to steer the pattern in a certain direction. By predicting the proper feeding phases of the antenna array elements, the neural network with the backpropagation technique and weighted hybrid gravitational search algorithm-particle swarm optimization approach is used to beam-steer the pattern. Furthermore, the DNN is applied to assign the required active subset elements for directing the main beam toward the desired direction. Several sample examples are provided to beam-steer the pattern in numerous directions to assess the correctness of the strategies.